Can Wearables Recognize Allergy Symptoms? Machine Learning with OpenEarable

What Makes This Thesis Interesting?

• Real-world application: Work on technology that could support future personalized health monitoring.
• Hands-on machine learning: Apply and evaluate algorithms using real multimodal wearable-sensor data.
• Flexible research focus: Shape the thesis around your interests, whether in classical ML, deep learning, signal processing, or wearable computing.
• Research environment: Gain experience working on an ongoing wearable-computing research project at TECO, KIT.

Depending on your interests and previous experience, your thesis could involve:

• Explore real wearable data: Analyze signals from motion sensors and ear-canal pressure sensors to understand how different activities and symptoms appear in the data.
• Develop machine-learning models: Implement and train models to recognize events such as coughing, sneezing, sniffing, or throat clearing.
• Experiment with different approaches: Compare established machine-learning methods, such as Random Forest or XGBoost, with time-series deep-learning methods.
• Investigate multimodal sensing: Explore whether combining multiple sensor signals improves recognition performance compared to individual sensors.
• Evaluate your models: Investigate how accurately the models recognize events and how well they generalize to previously unseen users.

You will work with existing datasets and a research prototype, allowing you to focus on the machine-learning and data-analysis challenges.

The scope will be adapted to your degree and interests.

For a Bachelor's thesis, the focus can be on implementing and evaluating established machine-learning methods for symptom recognition.

For a Master's thesis, the work can explore more advanced topics such as multimodal sensor fusion, anomaly detection, transfer learning, or generalization across users.

What should you bring:

• You study Computer Science, Electrical Engineering, Data Science, Medical Engineering, or a related subject.
• You have basic Python programming skills.
• You have some familiarity with machine learning or an interest in applying it to real sensor data.
• You enjoy experimenting, analyzing results, and solving technical problems.
• You can communicate in English.